日本語

Nicomachos Food Graph　v0.3

Food dataAI can answerwithout guessing.

NFG is an open food standard for holding dish, ingredient and allergen information as data the owner has verified. It is defined by NGraph Inc. and implemented in OMISEAI.

A pig and a chicken look at a fried skewer and wonder: pork or chicken? You cannot tell what is inside just by looking

AI answers about foodby guessing.

AI fills in ingredients and allergens with plausible guesses.

In a restaurant, one wrong answer can hurt someone.

NFG keeps owner-verified information as correct data (VAD), and AI answers only from that.

VADVerified Authentic DataCorrect data the owner has verified

Packaged food has the JAN code.Restaurants need a shared format too.

We collect dish information the owner has verified and structure it, working toward a food standard AI can refer to.

  • Menu✓ Owner-verified
  • Dishes✓ Owner-verified
  • Ingredients & allergens✓ Owner-verified
  • Restaurant✓ Owner-verified
NFGFood structuring standard

AI refers to it →

  • Generative AI and chat
  • Search and answer engines
  • Recommendations
  • Multilingual and overseas services

AI drafts,people verify.

We call this human-in-the-loop (HITL).

  1. 01AI draftsEstimates ingredients, cooking methods and allergens from menu photos and text
  2. 02The owner verifiesFixes what is wrong and approves what is right
  3. 03It becomes correct dataowner_verified, confidence 0.95 or higher
  4. 04AI answersAnswers only from verified data and shows it as owner-verified
Anything unclear goes back to the owner as a question

Every piece of datacarries its source.

Each item has where it came from (vadSource) and how certain it is (confidenceScore 0.0–1.0).

ai_inferredAI estimateInferred from photos or text. Not used in answers; sent to the owner to check
third_partyThird partyReviews, guidebooks and similar
official_publishedOfficially publishedThe restaurant's site or official menu
owner_verifiedOwner-verified (VAD)Correct data the owner has checked. This is what AI answers from
Structure

Ingredients, cooking methods, allergens, dietary restrictions (labeling rules in 5 countries)

Story

Taste, how to eat it, pairings and the dish's background

Structure for machines,story for people.　Both carry the same source and confidence. A dish is not fully described with only one of them.

OMISEAI putsthis standard to workin real restaurants.

Each dish card holds only verified information and answers guests in their own language. It is in live use at restaurants in Fukui and Kanazawa.

The same dish and ingredient dictionary translates restaurant websites into English, Chinese and Korean, so wording stays consistent.

Visit omiseai.com →
An OMISEAI dish card answering from owner-verified information

Put your food datainto NFG.

Restaurants

Send photos of your menu and get multilingual menus and AI guest service.

Visit OMISEAI →Companies, local governments and developers

Talk to us about using the standard, preparing data, or building it into your system.

Contact us (free) →

For developers

Use it as a JSON-LD vocabulary that extends schema.org.

Namespace
https://nfg.ngraph.jp/schema/v0.3/
JSON-LD context
/schema/v0.3/context.jsonld
v0.2：/schema/v0.2/context.jsonld
Example (a fictional restaurant)
{
  "@context": [
    "https://schema.org",
    "https://nfg.ngraph.jp/schema/v0.3/context.jsonld"
  ],
  "@type": ["Restaurant", "nfg:SeafoodRestaurant"],
  "name": "Sakanaya (fictional)",
  "hasMenu": {
    "@type": "Menu",
    "hasMenuItem": [{
      "@type": ["MenuItem", "nfg:Dish"],
      "name": "Boiled snow crab",
      "nfg:mainIngredients": ["snow crab"],
      "nfg:scientificName": "Chionoecetes opilio",
      "nfg:seasonStart": "11-06",
      "nfg:seasonEnd": "03-20",
      "nfg:vadCertified": true,
      "nfg:vadSource": "owner_verified",
      "nfg:confidenceScore": 0.95
    }]
  }
}
Types
nfg:SeafoodRestaurantnfg:CocktailBarnfg:KappoRestaurantnfg:KaisekiRestaurant new in v0.3nfg:Dishnfg:Ingredient
Properties
  • nfg:vadCertifiedOwner-verified or not (boolean)
  • nfg:vadSourceSource of verification (4 values below)
  • nfg:confidenceScoreConfidence 0.0–1.0
  • nfg:culturalContextCultural background for international guests
  • nfg:cookingMethodsCooking methods (set)
  • nfg:dishCategoryDish category
  • nfg:mainIngredientsMain ingredients (set)
  • nfg:courseFlowCourse order (ordered list)new in v0.3
  • nfg:provenanceOrigin / sourcing
  • nfg:scientificNameScientific name (fish, crab, etc.)
  • nfg:habitatHabitat (e.g. 200–600 m depth)new in v0.3
  • nfg:certificationTagOrigin or quality tag (e.g. the yellow tag on Echizen crab)new in v0.3
  • nfg:seasonStartSeason start (MM-DD)
  • nfg:seasonEndSeason end (MM-DD)
vadSource values
ai_inferredthird_partyofficial_publishedowner_verified
Versioning

During 0.x, breaking changes increment the minor version. v0.3 adds four terms already used on live sites (KaisekiRestaurant, habitat, courseFlow, certificationTag). v0.2 remains available.

License: the vocabulary is planned to be CC BY 4.0 (not yet final).